The Amazon is the largest tropical rainforest in the world, and thus plays a major role on global water, energy, and carbon cycles. However, it is still unknown how the Amazon forest will respond to the ongoing changes in climate, especially droughts, which are expected to become more frequent. To help answering this question, in this thesis I developed and improved the representation of biophysical processes and photosynthesis in the Ecosystem Demography model (ED-2.2), an individual-based land ecosystem model. I also evaluated the model biophysics against multiple data sets for multiple forest and savannah sites in tropical South America. Results of this comparison showed that ED-2.2 is able to represent the radiation and water cycles, but exaggerates heterotrophic respiration seasonality. Also, the model generally predicted correct distribution of biomass across different areas, although it overestimated biomass in subtropical savannahs. / Earth and Planetary Sciences

Just as the carbon dioxide observations that form the Keeling curve revolutionized the study of the global carbon cycle, free and open access to all available Landsat imagery is fundamentally changing how the Landsat record is being used to study ecosystems and ecological dynamics. This dissertation advances the use of Landsat time series for visualization, classification, and detection of changes in terrestrial ecological processes. More specifically, it includes new examples of how complex ecological patterns manifest in time series of Landsat observations, as well as novel approaches for detecting and quantifying these patterns. Exploration of the complexity of spectral-temporal patterns in the Landsat record reveals both seasonal variability and longer-term trajectories difficult to characterize using conventional bi-temporal or even annual observations. These examples provide empirical evidence of hypothetical ecosystem response functions proposed by Kennedy et al. (2014). Quantifying observed seasonal and phenological differences in the spectral reflectance of Massachusetts’ forest communities by combining existing harmonic curve fitting and phenology detection algorithms produces stable feature sets that consistently out-performed more traditional approaches for detailed forest type classification. This study addresses the current lack of species-level forest data at Landsat resolutions, demonstrating the advantages of spectral-temporal features as classification inputs. Development of a targeted change detection method using transformations of time series data improves spatial and temporal information on the occurrence of flood events in landscapes actively modified by recovering North American beaver (Castor canadensis) populations. These results indicate the utility of the Landsat record for the study of species-habitat relationships, even in complex wetland environments. Overall, this dissertation confirms the value of the Landsat archive as a continuous record of terrestrial ecosystem state and dynamics. Given the global coverage of remote sensing datasets, the time series visualization and analysis approaches presented here can be extended to other areas. These approaches will also be improved by more frequent collection of moderate resolution imagery, as planned by the Landsat and Sentinel-2 programs. In the modern era of global environmental change, use of the Landsat spectral-temporal domain presents new and exciting opportunities for the long-term large-scale study of ecosystem extent, composition, condition, and change.

Manmade radionuclides have been introduced to the environment for almost a century. The main source has been the nuclear weapons testing programmes, but accidental releases from the nuclear power production industries have also contributed. The risk to humans from potential releases from nuclear facilities is evaluated in safety assessments. Essential components of these assessments are exposure models, which estimate the transport of radionuclides in the environment, the uptake in biota, and transfer to humans. Recently, there has been a growing concern for radiological protection of the whole environment, not only humans, and a first attempt has been to employ model approaches based on stylised environments and transfer functions to biota based exclusively on bioconcentration factors. They are generally of a non-mechanistic nature and involve no knowledge of the actual processes involved, which is a severe limitation when assessing real ecosystems. The research presented in this thesis attempts to introduce a methodology for modelling exposure of biota that is based on systems ecological theories and concepts. All presented papers concern bioaccumulation and circulation of radionuclides in coastal areas of the Baltic Sea, which is a sea surrounded by several nuclear power plants, waste repositories and reprocessing facilities. Paper I illustrates how an ecosystem model can be used to predict the fate of C-14 in a bay, and to explore the influence of uptake route and water exchange on the concentrations in biota. Due to the longevity of many radionuclides, time spans of thousands of years need to be considered in assessments of nuclear waste facilities. In Paper II, the methodological problems associated with these long timescales are discussed and a new modelling approach is proposed. An extension and generalisation of the C-14 flow model into a generic model for other radionuclides is described and tested in Paper III. This paper also explores the importance of three radionuclide specific mechanisms (plant uptake, excretion and adsorption to organic surfaces) for the concentrations in biota. In Paper IV, the bioaccumulation kinetics of three radionuclides in three key benthic species of the Baltic Sea is studied experimentally. Paper V considers remobilisation and redistribution of sediment-associated radionuclides due to biological mixing, in a microcosm study. The findings in this thesis show both that it was possible to use an ecosystem approach to assess the exposure to biota, and that this approach can handle many of the problems identified in the use of traditional exposure models for radionuclides. To conclude, frameworks for the protection of the environment from ionising radiation would benefit from implementing methodologies based on ecologically sound principles and modelling techniques.

Cette thèse a pour objectif d’identifier les trajectoires plurimillénaires des écosystèmes de Corse à basse et moyenne altitude sous la triple influence des forçages climatiques, eustatiques et anthropiques à travers l’étude des assemblages d’insectes fossiles et des données paléobotaniques disponibles ou nouvellement acquises. Cette approche multidisciplinaire a été appliquée sur trois séquences sédimentaires prélevées à l’étang du Greco sur l’île Cavallo, au marais de Cannuta et à la tourbière de Bagliettu.Les résultats obtenus révèlent des changements majeurs dans la structuration du couvert végétal. Des signes d’ouverture du paysage apparaissent sur Cavallo et à Cannuta dès 5500-5000 cal. BP. Ils coïncident avec des indices d’activités des sociétés insulaires, lesquelles s’intensifient après 3000 cal. BP et plus particulièrement depuis 1000 ans à Bagliettu. Notre étude suggère que la sensibilité et la réponse des zones humides littorales face à la remontée relative du niveau marin dépend du contexte géomorphologique. Sur l’île Cavallo où le relief est faible, la transgression marine induit une augmentation de la salinité dans l’étang du Greco vers 3700 cal. BP, qui provoque une chute irréversible de la diversité des coléoptères aquatiques et hygrophiles. Au marais de Cannuta, où l’érosion des versants a favorisé la progradation de la plaine alluviale, notre étude montre une diminution de l’influence marine après 5000 cal. BP et une diversification de cette même entomofaune à partir de 1200 cal. BP. À Bagliettu, nous avons démontré l’importance des changements hydro-morphologiques des rivières dans la dynamique des écosystèmes tourbeux attenants et de leur entomofaune. / This thesis aims to define the multi-millennial trajectories of Corsican ecosystems under the climatic, eustatic and anthropic factors via the study of fossil insect assemblages and previous or new palaeobotanical data. This multidisciplinary approach was carried out at low and middle elevation on three sedimentary archives from the Greco pond on Cavallo Island, the Cannuta marsh and the Bagliettu peat-bog. The results reveal major changes in the structure of the vegetation cover over the last 7000 years. Signs of more open landscapes occurred on Cavallo and in Cannuta from 5500-5000 cal. BP. They coincide with bio-markers of insular societies activities, which increased from 3000 cal. BP onward and more particularly during the last 1000 years at Bagliettu.This study also suggests that the sensitivity and the response of coastal wetlands to the relative sea-level rise depend on the geomorphological context. On the low-lying Cavallo Island, marine transgression induced an increase of salt conditions in the Greco pond from 3700 cal. BP, which in turn, caused an irreversible loss of aquatic and hygrophilous beetle diversity. In Cannuta marsh, erosion on catchment slopes favoured the progradation of the coastal floodplain and the results reveal a progressive decrease of the marine influence from 5000 cal. BP as well as a diversification of this entomofauna over the last 1200 years. In Bagliettu, the results show how hydro-morphological changes of the rivers impact the dynamics of adjacent peaty ecosystems and beetle diversity.

Records of abundance of salmon that return to their natal spawning stream (escapements) are important indices that can assist with monitoring, conservation, and management of a salmon population over time. On their own, however these data reveal very little about the habitat, ecosystem and human communities that salmon encounter on their journey from freshwater to sea and back again. This research examines monitoring protocols for Goldstream River salmon stocks (coho, chinook and chum Pacific salmon). It includes and reaches beyond biostatistics from stream surveys to gauge First Nations’ artisanal fishing activities at Goldstream River and Saanich Inlet as well as their commercial chum fishing endeavours in Saanich Inlet on south Vancouver Island, British Columbia. Methods included summations of major themes from interviews on traditional ecological knowledge (TEK) shared by local Saanich First Nation fishers whose families have lived in the communities around Goldstream River and Saanich Inlet for more than 200 years. Analyses of Goldstream salmon escapements for the period 1932 to 2004 and native harvest statistics of chum caught from Saanich Inlet between 1982 and 2004 are integrated with results from analysis of TEK research undertaken for this project. Key recommendations arising from the results of this research are: stream habitat restoration in response to loss and degradation of salmon-bearing streams; modification of stream survey procedures to measure for morphological and physiological attributes including indicators of the health of Goldstream salmon; monitoring and eliminating sources of pollution to Saanich Inlet waters; implementing precautionary measures to ensure that overfishing of Goldstream salmon and shrimp in Saanich Inlet does not recur; and safeguarding naturally abundant Goldstream chum populations at the river. Under current management of the Goldstream chum fishery, the maximum carrying capacity (K) or target escapement of chum that the Goldstream River spawning grounds sustain is 15,000. Based on population assessments as well as physiography and ecosystem dynamics, I infer that Goldstream River’s K for its natural chum population is between ~16,000 and 18,000; ~1,500 for the mixed stocks of natural and hatchery enhanced coho; and ~50 for chinook (based on the river’s naturally occurring populations between 1932 and 1973) or ~385 enhanced chinook (based on the returning population from 1975 to 2002 since hatchery enhancement took place). A co-management relationship exists between Fisheries and Oceans Canada (DFO) resource managers and the Saanich First Nations bands (Saanich Tribal Fisheries councilors). Improvements to communication, collaboration and information sharing between DFO resource managers, Goldstream hatchery operators and Saanich First Nations with regards to decisions made about Goldstream salmon stocks are, however, necessary. In this thesis, I propose a model with recommendations for compatible fisheries management goals and techniques including adaptive management and ecosystem-based management to address this problem.

Records of abundance of salmon that return to their natal spawning stream (escapements) are important indices that can assist with monitoring, conservation, and management of a salmon population over time. On their own, however these data reveal very little about the habitat, ecosystem and human communities that salmon encounter on their journey from freshwater to sea and back again. This research examines monitoring protocols for Goldstream River salmon stocks (coho, chinook and chum Pacific salmon). It includes and reaches beyond biostatistics from stream surveys to gauge First Nations’ artisanal fishing activities at Goldstream River and Saanich Inlet as well as their commercial chum fishing endeavours in Saanich Inlet on south Vancouver Island, British Columbia. Methods included summations of major themes from interviews on traditional ecological knowledge (TEK) shared by local Saanich First Nation fishers whose families have lived in the communities around Goldstream River and Saanich Inlet for more than 200 years. Analyses of Goldstream salmon escapements for the period 1932 to 2004 and native harvest statistics of chum caught from Saanich Inlet between 1982 and 2004 are integrated with results from analysis of TEK research undertaken for this project. Key recommendations arising from the results of this research are: stream habitat restoration in response to loss and degradation of salmon-bearing streams; modification of stream survey procedures to measure for morphological and physiological attributes including indicators of the health of Goldstream salmon; monitoring and eliminating sources of pollution to Saanich Inlet waters; implementing precautionary measures to ensure that overfishing of Goldstream salmon and shrimp in Saanich Inlet does not recur; and safeguarding naturally abundant Goldstream chum populations at the river. Under current management of the Goldstream chum fishery, the maximum carrying capacity (K) or target escapement of chum that the Goldstream River spawning grounds sustain is 15,000. Based on population assessments as well as physiography and ecosystem dynamics, I infer that Goldstream River’s K for its natural chum population is between ~16,000 and 18,000; ~1,500 for the mixed stocks of natural and hatchery enhanced coho; and ~50 for chinook (based on the river’s naturally occurring populations between 1932 and 1973) or ~385 enhanced chinook (based on the returning population from 1975 to 2002 since hatchery enhancement took place). A co-management relationship exists between Fisheries and Oceans Canada (DFO) resource managers and the Saanich First Nations bands (Saanich Tribal Fisheries councilors). Improvements to communication, collaboration and information sharing between DFO resource managers, Goldstream hatchery operators and Saanich First Nations with regards to decisions made about Goldstream salmon stocks are, however, necessary. In this thesis, I propose a model with recommendations for compatible fisheries management goals and techniques including adaptive management and ecosystem-based management to address this problem.